Research And Implementation Of An Improved Recommendation Algorithm In Big Data Environment

In recent years,the rapid development of the Internet,which makes the information appear dramatic expansion.Many useful value is found in large amounts of information,but many people also paid a price.Recommended system is commonly used as a means for filtering the information.This means has drawn increasing attention.Collaborative filtering algorithm is simple,independent,considering only the historical ratings data,it has become one of the most frequent recommendation algorithm.However,it also has some drawbacks,such as cold start,data sparseness,difficult to expand and the like.Meanwhile,with the development of many new technologies,the order of the data has reached PB ZB and the like.Current society has entered the era of big data.The number of items and users has been constantly increasing,Stand-alone system of computing time,storage space,which have become an important factor affecting the performance of the recommendation.How to improve the traditional recommendation algorithm in a big data environment,which has become a serious problem.In order to achieve better results,resolve the foregoing problems,the paper analyzes the traditional classical recommendation algorithm and several improved algorithms,then some problems exist in the traditional recommendation algorithm,cold start,sparse data,etc.,this paper proposes a modified recommendation algorithm,it is the combination of demographic information and user trust mechanism.The final results showed that: this method can not only improve the accuracy and demographic information through the trust mechanism.They considered the impact of multiple factors,greatly ease the cold start and the data sparseness problem.In addition,in big data environments,when faced with massive data,the system may appear difficult to extend computing performance and other issues,In this paper,the algorithm(Improved recommendation algorithm)parallel processing,then,for the relevant comparison test.Experimental results show that the effect of parallelism in this article to improve the ease of data sparsity recommendation algorithm to solve the problem of cold start and improve the recommendation accuracy.The scalability and computational efficiency have preferred.Finally,this paper designs an online catalog recommendation system based on improved recommendation algorithm.It also verifies the practicality and effectiveness of the algorithm.Through the system requirements analysis,system design and describes the framework and process,which uses a distributed Hadoop framework,JavaWeb technology and MySQL database and show the recommended results.